Rundeck GenAI Job Creator [SaaS][product design][generative AI][B2B]

Rundeck's core workflow involves building automated jobs — but job creation was limited to power users. The GenAI Job Creator changed that, letting users type a prompt and generate a ready-to-go workflow automatically. It's the first completed GenAI project in Rundeck's history.

Impact:First GenAI project shipped in Rundeck · Lowered job creation barrier for non-expert users

Role:Product Design, Design Exploration, Mentorship

Collaborators:Jake Cohen (PM), Jenny, Will Su

Why

Building automated jobs in Rundeck required deep platform knowledge — general users rarely created jobs because the complexity and prerequisite knowledge were too high.

How

A GenAI-powered job creator that lets users type a natural language prompt and generates a ready-to-go workflow automatically — the first GenAI feature completed within Rundeck.

Impact

First GenAI project shipped in Rundeck · Lowered job creation barrier for non-expert users

I designed the first GenAI feature completed within Rundeck — making automated job creation as simple as typing a prompt.

I mapped the full user flow — from signup and job list entry through GenAI parameter selection: category, character limit, and prompt input.

User flow mapping the GenAI job creation path from signup through the jobs list into the creation UI, with branches for category selection, character limit enforcement, and prompt input.
User flow mapping the GenAI job creation path from signup through the jobs list into the creation UI, with branches for category selection, character limit enforcement, and prompt input.

I accounted for every outcome after generation — loading states, failure handling, and three success paths.

Continuation of the flow covering all post-generation outcomes — loading feedback, three failure states (token limit, service busy, off-topic prompt), and three success paths (save job, add to library, edit classically).
Continuation of the flow covering all post-generation outcomes — loading feedback, three failure states (token limit, service busy, off-topic prompt), and three success paths (save job, add to library, edit classically).

An early split-panel exploration separated input from output — a category dropdown and prompt field on the left, the generated job definition rendered on the right.

Early split-panel exploration with a category dropdown and prompt field on the left, and the generated Rundeck job definition with steps, options, and UUID on the right.
Early split-panel exploration with a category dropdown and prompt field on the left, and the generated Rundeck job definition with steps, options, and UUID on the right.

I explored a chat-like interface anchored inside the existing job creation tabs, using curated categories as onboarding and a prompt input as the only required action.

Wireframe of the GenAI tab embedded in Rundeck's job creation flow — category chips and example prompts guide users toward a prompt input, with the generated job definition rendered inline on the right.
Wireframe of the GenAI tab embedded in Rundeck's job creation flow — category chips and example prompts guide users toward a prompt input, with the generated job definition rendered inline on the right.

Onboarding disguised as navigation — category chips orient users without a blank slate, and the job preview breaks the output into numbered, scannable steps.

Annotated final design showing the prompt entry view with category-based onboarding and example prompts, and the job preview view with numbered steps surfaced from the generated runbook.
Annotated final design showing the prompt entry view with category-based onboarding and example prompts, and the job preview view with numbered steps surfaced from the generated runbook.

The final interface embedded directly into Rundeck’s existing UI — categories and example prompts guide users, with the selected prompt populated in the input field.

Final GenAI Job Creator UI with curated category chips and contextual example prompts reducing the blank-slate problem and guiding users toward a first generation.
Final GenAI Job Creator UI with curated category chips and contextual example prompts reducing the blank-slate problem and guiding users toward a first generation.

After submitting a prompt, a progress bar and status text give users real-time feedback while the job is being generated.

Generation loading state with the submitted prompt echoed back and a progress bar showing "Determining job steps…37%" communicating that generation is actively in progress.
Generation loading state with the submitted prompt echoed back and a progress bar showing "Determining job steps…37%" communicating that generation is actively in progress.

The generated job surfaces as a named preview with discrete, expandable steps — rendered in Rundeck’s native UI patterns so the output feels familiar, not foreign.

Generated job preview showing "Export MySQL Query to S3 as CSV" broken into three numbered steps — SQL query, CSV export, and S3 upload — rendered in Rundeck's native step UI with a "Create this job" CTA.
Generated job preview showing "Export MySQL Query to S3 as CSV" broken into three numbered steps — SQL query, CSV export, and S3 upload — rendered in Rundeck's native step UI with a "Create this job" CTA.

It shipped as the first completed GenAI project in Rundeck’s history — lowering the job creation barrier for non-expert users who previously couldn’t get started without deep platform knowledge.